Take-Two CEO says AI efficiency may not cut game development costs

Zelnick describes himself as fully supportive of AI. His reservation concerns what the technology can originate: he believes it currently cannot provide the genuine originality and surprise central to blockbuster games. His position leaves substantial room for AI assistance while setting a limit on what he thinks that assistance can accomplish independently.

He bases that judgment on AI's dependence on existing data. Zelnick characterizes it as a combination of large datasets, computing power and large language models, and emphasizes that the data reflects the past. That is why he questions claims that AI represents an entirely new creative force: in his assessment, drawing on previous material does not itself supply the unexpected ideas a hit needs.

At the time of the discussion, GTA5 had sold more than 200 million copies worldwide since its 2013 release, providing a concrete measure of the commercial success under discussion. Zelnick accepts that AI can produce something resembling GTA, but argues that imitations struggle to sell. Resemblance, in that argument, is a different achievement from creating a game capable of comparable commercial success.

Asset generation is where he sees AI's strongest current application. Yet he separates the production of those assets from the creation of a hit: a successful game requires more than assembling material. Producing assets more easily still leaves the question of an entire game's originality, which is the creative problem he emphasizes.

The competitive implications matter to some investors. Their concern is that increasingly capable AI tools could lower the barriers to game development and expose established publishers such as Take-Two to greater competition. The question extends beyond whether existing teams can work faster: easier production could also give potential challengers more accessible ways to make games.

Zelnick disputes that assessment by pointing to the tools developers already have. He argues that game development tools became widely available years ago, while ideas remain the scarce resource. In his view, wider access to production technology does not automatically remove the creative obstacle that separates being able to build a game from developing a distinctive hit.

In earlier remarks on AI, Zelnick also discussed Take-Two's approach: the company had allowed employees to use tools including Claude and Gemini to assist their work. This example fits the openness he has expressed toward the technology. He supports developers using those tools while remaining reserved about AI independently originating an entire GTA-scale hit.

His earlier prediction about costs follows a separate mechanism: easier development can encourage higher goals. Zelnick expects AI efficiency gains may raise expectations for content scale and quality rather than substantially lower AAA development costs. Under that scenario, individual tasks become easier, but a more ambitious overall project creates additional work. The savings could be absorbed by greater development demands instead of reducing the total production bill.